Date post: | 01-Dec-2023 |
Category: |
Documents |
Upload: | independent |
View: | 0 times |
Download: | 0 times |
THE COMBINATION OF GEOMATIC APPROACHES AND OPERATIONAL
MODAL ANALYSIS TO IMPROVE CALIBRATION OF FINITE ELEMENT
MODELS: A CASE OF STUDY IN SAINT TORCATO CHURCH (GUIMARÃES,
PORTUGAL).
Luis Javier Sánchez-Aparicio1*
, Belén Riveiro2, Diego Gonzalez-Aguilera
1, Luís F. Ramos
3
1Department of Land and Cartographic Engineering. University of Salamanca,
High Polytechnic School of Avila, Hornos Caleros, 50, 05003, Avila (Spain)
Tlf: +34 920353500; Fax: +34 920353501
2Department of Material Engineering, Applied Mechanics and Construction.
School of Industrial Engineering. University of Vigo. Vigo, (Spain)
Tel.:+34 986 813 661; fax:+34 986 811 924.
3ISISE, Departament of Civil Engineering, University of Minho, Guimarães (Portugal)
Tlf: +351 253510200; Fax: +351 253510217; [email protected]
*Corresponding author: Tel.:+34 920353500; ; fax:+34 920353501
E-mail address: [email protected]
Abstract
This paper present a set of procedures based on laser scanning, photogrammetry (Structure
from Motion) and operational modal analysis in order to obtain accurate numeric models which
allows identigying architectural complications that arise in historical buildings. In addition, the
method includes tools that facilitate building-damage monitoring tasks. All of these aimed to
obtain robust basis for numerical analysis of the actual behavior and monitoring task.
This case study seeks to validate said methodologies, using as an example the case of Saint
Torcato Church, located in Guimãres, Portugal.
Abbreviations: UAV, Unmanned Aerial Vehicle; SfM, Structure from Motion PW,
Photogrammetry Workbench; MAC, Modal Assurance Criterion; DR Douglas-Reid Method,
ICP Iterative Closest Point; NURBS Non-Uniform Rational B-Splines.
Keywords: Structure from motion; Laser scanner surveying; Finite Element Model; Modal
updating; Masonry structures; Orthoimage; Damage Survey; CAD modelling.
1. Introduction
The conservation of historic buildings requires understanding their structural behaviour, and
consequently: (i) Their boundary conditions, (ii) The characteristics of the constitutive materials
(iii) The origin of the damage that the building suffers and (iv) Their vulnerability [1]. Therefore
the creation of accurate numerical models is imperative in order to obtain adequate restoration
systems.
Masonry walls are very common in the vast majority of existing monuments. Cracked elements,
associated with different events (settlements and/or excessive displacement loadings) are a
common problem that reduces the service life of these structures [2]. The fracture phenomena
(cracks) are caused by the masonry’s high brittleness to tensile stresses. Furthermore, the
structural behaviour is highly dependent of the structural geometry. This is why four conditions
are required to carried out proper analysis: (i) Having a complete and accurate geometric
characterization of the structure; (ii) Knowing the material´s mechanical properties (iii)
Characterizing all the loads acting in the structure; and (iv) Providing numerical models that
correctly simulate the characteristic behaviour of the structure (non-linear material behaviour,
ground settlement, contact between bricks, etc.).
Modern restoration techniques for built heritage are characterized by minimal intervention,
compatibility, durability and reversibility [3]. Identifying and to monitoring the pathological
condition of the building plays a key role in understanding the current behaviour of the structure
and the choice of restoration methods to be accomplished [4].
Traditional measuring methods often had a significant dependence of the worker’s skills and
they normally have associated high time cost. These methods were replaced by direct
interpretations done over the building plans (design models) [5-7]. The constant progress of the
numeric method of finite elements and computer processing allows the generation of
increasingly complex geometric models; that is why it is more and more imperative the
necessity of relying on sensors capable to provide massive detailed data and features for the
model. Is in this field where geomatic sensors like terrestrial laser scanner [8, 9] or digital
camera [10] have acquired important roles, due to the capacity of acquire accurate geometrical
information needed by the numerical models.
In the present paper, the proposed methodology for data acquisition combines and enhances the
laser scanning and digital camera system providing, beside the characteristics defined above, the
versatility of adaptation to different infrastructures. All of this within a single application, a
hybrid point-cloud, which greatly eases the preparation of geometrically precise numerical
models that also serve as a basis for the monitoring of the structure through the analysis of,
either the point-cloud or by the analysis of the obtained orthoimages.
The paper relies on the application of the proposed methodology on a case study, St. Torcato
Church, close to the city of Guimarães, Portugal [1]. This historical construction has moderate
to severe damage and needs to be strengthened. The methodology has been carried out to
upgrade and calibrate the finite element model using a global dynamic identification, including
crack and geometric improvement, in order to complement with the static and dynamic
monitoring system and a future numerical analysis. All this will be made in order to obtain the
current stress state of the building and asses the effectiveness of subsequent restoration
mechanism that aim at stabilizing the damage.
Within this context, this article attempts to demonstrate a methodology for data acquisition and
processing and it is organized in the following way: Section 1 is the introduction; Section 2
presents the instruments for data acquisition which are the laser scanner and the set formed by
the UAV and digital camera; Section 3 wherein the methodology for obtaining the hybrid point-
cloud and the calibration of the model is shown; Section 4 the data obtained by the laser scanner
and the digital camera sensors is analysed separately, for the presented case study and for the
potential of the hybrid point-cloud (this applies not only to the numerical finite-element analysis
but also to damage analysis monitoring); and finally in Section 5, the conclusions are drawn.
2. Materials and methods
As described in Section 1, the aim of this methodology is to generate precise finite-element
numerical models for subsequent structural analysis. These must be precise, in terms of
geometry accuracy and they must contain the necessary data to monitor and track the evolution
of damages in the structure. All this within an accurate georeferenced framework and with non-
intrusive sensors as the main source.
2.1. Laser scanner system: the terrestrial laser scanner
Currently the terrestrial laser scanning system has acquired great relevance by offering a wide
range of advantages; one of the greatest one is the acquisition of non-contact three-dimensional
geometry of the analyzed surface, preventing any disruption and allows to accurately capture
geometry, providing a high density of data (millions of points) [11]. This feature includes that it
does no dependency on specific lighting conditions [12]. It is therefore the combination of
accuracy, speed and range of measures that has placed the system as the most powerful tool for
three-dimensional modeling and reconstruction of monuments [13, 14].
In order to establish a valid methodology to make more accurate finite element models, two
different laser scanning systems have been analyzed based on different measurement principles
(Table 1.): (i) Riegl LMS-Z390i based on time of flight principle (ii) Faro Focus 3D based on
the phase shift principle. For details on these measurement principles, refer to [15].
Riegl LMS Z-390i Faro Focus 3D
Measurement principle Time of flight Phase shift
Wavelength 1550 nm 905 nm
Measurement range 1-400 m 0.6-120 m
Accuracy nominal value
6 mm a 50 m in specific lighting
and reflectance conditions.
2 mm a 25m in specific lighting
and reflectance conditions.
Field of view
3600 horizontal
800 vertical
3600 horizontal
3050 vertical
Capture rate 11000 points/sec 122.000/976.000 points/sec
Beam divergence 0.3 mrad 0.19 mrad
Table 1: Comparison of technical specifications between laser scanner system Rielg LMS Z-
390i and Faro Focus 3D.
2.2. Imaging System: UAV and “Structure from Motion”
While the laser scanning system allows fast capture and processing of data, it has some
drawbacks such as the difficulty for transport and the restriction of stationing in certain elevated
places inside historic buildings, these places often are critical. Therefore, it is necessary to use
additional platforms and sensors capable of providing accurate data from any position; for this
the onboard digital camera on an Unmanned Aerial Vehicle (UAV) platform is used (Fig. 1).
Fig.1. Image of the UAV platform and the digital camera (Left). Image taken during the data
collection of the point cloud SfM (Right).
The chosen photogrammetric platform was designed by Roca et al [16]. It is made of aluminum
and carbon fiber, and comprises a total of eight MK-3638 SLOWFLY APC propeller motors
controlled by a central 12x3.8 Brushless Control V2.0 that can manage separately the rotational
speed of each of the motors. All of this provides the system with great stability and robustness
against failure in flight.
In addition to this platform, a low-cost sensor, a Canon EOS 450D digital camera that had been
previously geometrically calibrated (Table 2), and a Canon EF 20mm wide-angle lens were
assembled. The wide-angle lens is meant to minimize the amount of images taken.
Parameter Value
Sensor size
W= 22.2425mm
H=14.8336mm
Principal point
Xp=10.8716mm
Yp=7.4449mm
Focal length f=20.4222mm
Radial distortion
K1= 2.157 e-004mm
K2= -4.189 e-007mm
K3=0mm
Tangential
distortion
P1= 4.321 e-005mm
P2= -1.003 e-005mm
Table 2: Canon EOS 450D digital camera geometric calibration settings.
In recent years photogrammetric data processing systems (SfM) have taken a great relevance;
they are able to include into their structure the advantages of computer vision (automation and
flexibility) and those of photogrammetry (accuracy and reliability) [17] in order to obtain dense
three-dimensional models (point clouds) that can compete in accuracy with the laser scanner
system [18]. Within this field highlights the Photogrammetry Workbench (PW) software, which
implements the “Structure from Motion” system, ensuring automation (in the transformation of
2-D images to 3-D point clouds), flexibility (by allowing work with any type of camera,
calibrated and non-calibrated) and quality (to ensure precision and quite acceptable resolutions).
3. Methodology
3.1. Generating the CAD model and its integration with finite elements
The early stage in the laser scanning and the Structure from Motion (SfM) data processing, have
been omitted in this article, since our main interest is focused on establishing a robust
methodology that serves as a template for subsequent restoration actions. This template will be
based on the hybrid point cloud, which comes from the combination of data obtained from laser
scanning, the SfM and the analysis of the products that are obtained from them. For more details
about SfM flow see either [19]
Also is imperative to building an accurate CAD model which allow us to evaluate the actual
behavior of the construction as a basis for the numerica analysis. However, at an early stage, the
point cloud provided by the laser scanning and the SfM present superabundant information in
different coordinate systems. As a result, this data is not suitable for CAD model building.
Following a semi-automatic method that allows adapting the point cloud to an accurate and
suitable CAD model for numerical analysis is presented (Fig.2).
This methodology requires a multi-phase post-processing that involves three main steps: (i)
Data fusion at the same coordinate system through registration algorithms; (ii) Point cloud
resampling and (iii) Point cloud simplification (removing certain architectural details without
relevance) and parameterization for CAD model conversion.
Fig. 2 Workflow for the proposed methodology.
3.1.1. Hybrid point cloud registration
A complete documentation of historical buildings requires the use of multiple point cloud data
set. Is a requirement therefore to place all of these point clouds in the same coordinate system in
order to be processed together.
The proposed methodology is based on a registration system coarse to fine. In an initial stage a
point cloud pair-wise registration is carried out. This step takes as a base the ICP (Iterative
Closest Point) algorithm[20], that minimize the difference between two points clouds, requiring
a total of n-1 alignments, where n is the number of point clouds.
Using a pair-wise registration causes an error propagation along the registration of all the point
cloud scans. In order to minimize this error accumulation a global registration, based on
Generalized Procrustes Analysis[21], was used considering the pair-wise registration,
previously made, as the rough registration needed.
3.1.2. Hybrid point cloud resampling and CAD conversion
Traditionally the step procedure from the raw point cloud to the CAD model could be made
through three different approaches [22]: (i) Orthogonal views; (ii) Sections applied along
directions and over the mesh and (iii) Non-Uniform Rational B-Splines(NURBS) generated
from the mesh. The two first approaches require a high manual work made by the user, whereas
the NURBS approach demands high computational cost.
In this article an alternative and semi-automatic approach is presented, which combines NURBS
and parametric shapes approximations with the addition of a segmentation process described
below, in order to build a suitable CAD model for structural applications. While NURBS-based
method was used for complex shapes like (vault or domes), the parametric-based method was
used for the rest of the structure.
Once the registration procedure has been completed, the resulting point cloud needs to be
resampled (due to the high amount of data) in order to generate a suitable CAD model for the
numerical analysis. In this case several methodologies could be applied based on[23]: (i)
Principal Component Analysis; (ii) Quadric-Based Polygonal Surface Simplification; (iii)
Clustering methodologies and (iv) Radial Based Function. For the proposed methodology a
resampling based on curvature has been applied, in order to decimate flat surfaces without
losing detail in features areas. This curvature based resampling follows the next steps: (i)
Creating a local neighborhood of the analysis point; (ii) Local surface based on quadratic
approximation and (iii) Extraction of the normal and principal curvatures.
After that, the resulting point cloud is meshed, since the majority of segmentation procedures
are also performed over meshes. The segmentation process is performed by Functionally
Decomposed Surface Models[24]. Once the segmentation is done, the different surfaces created
are approximate to NURBS and parametric shapes. As a result a manageable CAD model is
generated and could be imported and used for a FEM package.
Additionally to the mentioned above, some relevant features like cracks, can be included into
the CAD model that defines more realistically the building’s behavior. It is sufficient to extract
the area of interest from the point cloud, either SfM or laser, to mesh that area and to
incorporate it into the CAD model (Fig. 2).
Fig. 3 Graphical description of the proposed methodology. From left to right: point cloud, mesh
model, segmentation model, CAD model with major cracks.
3.2. Crack recognition and characterization
Digital image analysis is a tool of great potential in the field of pathological characterization of
buildings. Several authors demonstrate the feasibility of this analysis to characterize either from
the terrestrial laser scanner [25, 26] or from the image captured by a digital camera [27, 28].
Besides offering robust tools for generating three-dimensional models from two-dimensional
data (digital image), PW can also obtain orthoimages on specific areas and specific levels,
providing the user with precise documentation from anywhere in the building.
Given the high density of points gathered during the three-dimensional reconstruction, the
production of orthoimages will require only two points to provide adequate scale, taken from
the laser point cloud and a reference plane, in order to run an orthographic projection. The pixel
size of this projection is calculated according to the density of the cloud of points, close to the
resolution of the initial images [29].
Given the fact that the obtained orthoimage stands as a product without geometric distortions
and in real scale, it is sufficient to directly measure on it and so obtaining crack characterization
(length and opening).
3.3. Finite element numerical model
The geometric accuracy and high level of detail provided by the laser scanning systems and
structure from motion, provide complex CAD models. In order to solve this geometric
complexity, for the discretization elements it is required: (i) high flexibility to adapt themselves
to the geometry and (ii) great compatibility with automatic meshing algorithms [7]. All this
makes the tetrahedral discretization elements with an isoparametric formulation the most
suitable for the meshing of complex CAD models.
3.4. Calibration of the finite element numerical model
The analytical results obtained from the calculation model are sensitive to material properties
and boundary conditions [30] thus making necessary to gather experimental data to optimize the
numerical model.
Among the different possibilities availables today for the implementation of in-situ tests on
historic buildings, experimental modal identification is the most popular method [1]. This
technique is a non-intrusive system with the capability to identify the global properties of the
structure. It allows to obtain vibration frequencies, damping coefficients and mode shape of
historic buildings, which may be related to various physical properties (Young modulus,
density, stiffness of connections between parts, etc.) wich makes possible to validate the
analytical models [31].
This publication builds on the data obtained from the accelerometers configuration adopted in
2009. The campaign counted a total of 35 measuring points with 9 sets spread throughout the
Church, for more details see [1] (Fig. 4).
Fig. 4. Scheme of the arrangement of the 35 accelerometers on the Church (Above). Mode
shape obtained from operational modal analysis (Bellow).
The basic objective of these methods is to improve the correlation between the experimental
data and those obtained from finite element model, through small changes in a group of model
parameters[32]. The criterion often used to assess the correlation is the MAC (Modal Assurance
Criterion), this being defined from the following formula (1)[33]:
𝑀𝐴𝐶𝑢,𝑑 =[(𝜑𝑖
𝑢)𝑇(𝜑𝑖
𝑑)]2
(𝜑𝑖𝑢)
𝑇(𝜑𝑖
𝑢)(𝜑𝑖𝑑)
𝑇(𝜑𝑖
𝑑) (1)
where 𝜑𝑖𝑢 y 𝜑𝑖
𝑑 correspond to the mode shape vector on experimental and numerical model
respectively for a vibration mode i.
As noted above, the goal is to minimize the existing differencies between the experimental
behaviour and numerical model, considering the experimental values as references. Used in
2007 for the calibration of the numerical model for the Monza Cathedral bell tower [34], the
methodology proposed by Douglas-Reid [35] (DR) can be used for calibration of finite element
numerical models. This method tries to minimize the difference between theoretical and
experimental parameters through the natural frequencies, or another modal parameter, using the
following approach:
𝑅𝑖𝐹𝐸(𝑋1, 𝑋2, … , 𝑋𝑛) = ∑ [𝐴𝑖𝑘𝑋𝑘 + 𝐵𝑖𝑘𝑋𝑘
2]𝑁𝑘=1 + 𝐶𝑖 (2)
To solve these equations, a total of 2n+1 values are required to be calculated, taking into
account initial values, as well as lower and upper bounds for all updating parameters.
Using the methodology followed by [34], the next step consists of determing the modal
frequencies and minimize their difference according to the following objective function (3)(4):
𝐽 = ∑ 𝑤є2𝑚𝑖=1 (3)
є = 𝑓𝑖𝐸𝑋𝑃 − 𝑓𝑖
𝐹𝐸(𝑋1, 𝑋2, … , 𝑋𝑛) (4)
where J is the objetive fuction to be minimized, w are the weights considered through
engineering criterion and e represents the error function (difference between the frequency
obtained by operational modal analysis 𝑓𝑖𝑂𝑀𝐴 and numerical analysis𝑓𝑖
𝐹𝐸).
The main drawback of this methodology lies in the consideration of a unique modal parameter;
frequency. To obtain more accurate results the objective function must be modified, including
the MAC values(5):
𝐽 = 1/2 [𝑊𝑓 ∑ (𝑓𝑖,𝐹𝐸𝑀2 −𝑓𝑖,𝐸𝑋𝑃
2
𝑓𝑖,𝐸𝑋𝑃2 )
2𝑚𝑖=1 +𝑊𝑀 ∑ (1 −𝑀𝐴𝐶𝑖,𝐹𝐸𝑀)
2𝑚𝑖=1 ] (5)
where J is the objetive function to be minimized, Wf and Wm are the weights considered for the
frequency and vibration modes, f is the frequency and 𝑀𝐴𝐶 the modal assurance criterion
values, both values corresponding to the vibration mode i.
4. Experimental result and discussion
Located in the village of St. Torcato, within the municipality of Guimarães (northern Portugal),
the Church of St. Torcato is a clear example of historic building built in stone material, showing
moderate-severe structural damage made evident mostly by cracks in its main façade. Starting at
the entrance arch keystone, it extends through the rosette to the coronation, splitting this
element in two macroblocks [1] (Fig. 5).
Such crack increases the width along its development up to the roof. The movement in opposite
directions of the main façade towers due to the settlement suffered by the building is
remarkable, as well as some “chrusing” type cracks caused by the compressive stress
concentration resulting from eccentric loads originating in the towers.
Fig. 5. Representation of the possible structural failure collapse mechanisms of the Church of
Saint Torcato. It is possible to observe the formation of two macro-blocks.
Built in style “Neo-Manuelinio” the Church of Saint Torcato mix Classics, Gothic, Renaissance
and Romanesque elements in its extension [6].This gives it a special and complex aesthetic that
along its length, with a height of about 50 m in the towers, prevents effective tridimensional
data capture with laser scanner, topographic techniques or manual measurements [36]. The
binomial structure from motion and laser scanning is the ideal solution allowing abundant and
accurate three-dimensional data capture anywhere in the building.
The results obtained are hereby analyzed independently, according to the source sensor (laser
scanner or digital camera) and the resulting numerical model of the combination of these and
the calibration using operational modal analysis.
4.1. Terrestrial Laser Scanner
For the study we have considered the two most popular measuring systems for survey of
buildings and civil infrastructures: the laser time of flight and phase difference [25].
Multiple tests have been carried out in the exterior as well as in the interior of the Church. Since
the point-cloud is defined by densisty of points, the acquisition rate and range, the laser scanner
LMS Z-390i Rielg (based on time of flight) is considered to be the most suitable for data
capture in the exterior. Besides, it has a larger range compared to the Faro Focus 3D laser.
Indoors, the data acquisition speed of the Faro Focus 3D scanner (122000 points/sec) compared
to the speed of the Riegl LMS laser Z-390i (11000 points/sec), together with its portability
proved to be the most important advantages for gathering the data in the interior of the Church.
While in both cases the laser system provides a sufficient and suitable density of points to
accurately monitor deformations [37], the amount of data and distinctive features provided
might be insufficient fo the extraction and monitoring of cracks (for example, it does not record
texture) (Fig. 6).
Fig. 6. Settlement of the entrance vault detail (Left). Settlement of the entrance vault detail
(Right).
The final model had a total of 29 scans and 267.601.626 points: (i) 14 scans were done of the
outside with the Riegl LMS laser Z-390i, (ii) 3 interior shots were taken with the Riegl LMS
laser Z-390i and (iii) 12 interior scans were made with the Faro Focus 3D laser. However, the
top of the towers and the rooftop of the Church could not be modelled because no suitable
location was found for the laser to reach those areas. In addition, the data collection was
hampered by additional conditions: the excesive laser beam skew angle on the ledge, top of the
towers an the openings between the chapels preventing a complete and accurate characterization
of the building and of the critical areas. Thus requiring a complementary technique capable of
solving such weaknessess, UAV and SfM.
4.2. UAV and Structure from Motion
The point-clouds collected by laser scanning do not provide a sufficient amount of data for a
full geometric characterization of the outer shell: either the distance from scanner to object the
range is too large, the point of view is insufficient or the laser system cannot be placed in a
certain location, such as in the upper region of the façade beyond the central cornice.
In addition to the aforementioned, the obliqueness phenomenon must be considered. As shown
by authors such as [38, 39], this phenomenon is highly correlated with the value of uncertainty
in obtaining a point’s spatial coordinates. This phenomenon is of great importance in obtaining
accurate products and it is highlighted by this case study, where it is critical in certain areas,
such as in the inclination of the towers or in the cracking of the central façade.
All this requires a supplementary technology to the laser scanning, this is UAV + SfM; a non-
intrusive way of solving the problems described above through its great portability and ability to
collect data. Besides, it gathers an extra supply of analyzable features, which makes it possible
to get complete point-cloud models, which form the foundation for accurate and thorough CAD
models. These models profit from the features obtained from the hybrid point-cloud, such as
cracks (Fig. 7).
Fig. 7. Front elevation of the point cloud obtained through SfM technique and PW software
(Left). Detailed comparison between the laser scanner point cloud and the one obtained in PW
software (Right).
This model, generated through the describend technique, is comprised of a total of 398
photographs taken by UAV platform: (i) 273 photos of the main façade dividen in 3 vertical
strips ( 1 for each tower and one for the main façade) and (ii) 125 photos of the cracks on
chapels, also divided in 3 vertical strips. Alongside these photographs, 117 aditional shots were
taken without UAV platform (terrestrial photogrammetry): (i) 85 photos of the arches of the
chapels of the main nave and (ii) 32 photos at the level of the Church choir.
Both techniques, terrestrial laser scanner and SfM, complement to each other, and their
combination is the ideal solution for restoring built heritage. While the laser scanner provides
the system a set of precise, dense and fast capture data with which it is possible to monitor
structural movement and generate a CAD model suitable for numerical analysis, the UAV+ SfM
system combines with it perfectly supplying the geometric data of the areas that were
unreachable through the previous system. In addition, it characterizes completely the structure’s
pathological conditions by obtaining direct and georeferenced data.
However, the resulting data of both sensor show different coordinate system but the high
redundancy allows the creation of a single model applying the methodology explained in section
3, that finally is converted in to a valid CAD model of the building for the subsequent analysis.
An average value of XXXX with a standar deviaton of XXXX was found in the coarse
registration process. Later, in the fine registration process this values down to XXXX with a
standard deviation of XXXX. The final hybrid point cloud had a total of 40359060 points (that
represents a 10% of the original set).
4.3. Characterisation of structural pathologies
The characterization of cracks plays a fundamental role in structure monitoring in terms of
stability and safety. Traditionally such monitoring was conducted with graduated cards,
mechanical or electronic gauges, or LVDT ( linear variable differential transformer ). However,
this equipment has significant drawbacks [27]: (i) Firstly, there is a need for permanent plates,
that may become damaged or can be lost, (ii) They provide data only from certain points and
certain directions, (iii) The cracking is not directly measured; it is assumed that its activity is
correctly defined by the variation of the reference points. In addition to to this, some of these
methods strongly rely on temperature ( this is the case of electronic gauges).
Thanks to the combined use of the shown spatial-data capture techniques, it is possible not only
to obtain high density point-clouds and photorealistic textures (this is the case of UAV+SfM
system) but also high quality orthoimages in any position and on determided surfaces, thus
solving the problems described above. All supplemented with direct product georeferencing,
which makes perfectly viable to monitor the movement of the structure and the evolution of
damage that may arise.
The aforementioned methods combined with an accurate numerical model will comprise all the
necessary tools for sizing and evaluating the restoration system of the building.
The first damage inspection of the monument [6] was carried out in 1998. The inspection
indicated that the façade suffered structural damages, made evident by cracks running from its
bottom, in the keystone to the coronation. In addition, patologies are observed in the entrance
dome keystone under the choir, in the arcs are that make up different bays of the main nave and
in cracks on the side of the building.
It is on the main facade where the building shows a greater amount of these pathological
conditions, spread along as cracking and displacement on elements of arches and vaults (Fig.8).
Fig. 8. Results of the inspection for damages: Plan view at ground level (Left). Main façade
orthoimage inspection for damages (Right).
4.4. Geometrical CAD model
Made by the proposed methodology, the Saint Torcato CAD model has greater geometric
complexity than the one exposed by Lourenço and Ramos[6]. Within this geometric
improvement is remarkable a better characterization of the main façade and towers, including
architectural details over the balcony and along the towers.
Since most of the façade shows structural pathological conditions, it is therefore expected that
the dynamic response of the structure will be influenced in part by such cracking. The high
correlation between the CAD model and the actual photogrammetric point cloud allows to
incorporate these characteristics into the CAD model directly, only requiring the meshing of the
area under study. Also this model, presents different wall tickness, increasing the model realism.
(Fig. 9).
Fig. 9. Isometrics views of the initial geometric model (Left) and the one generated by the
proposed methodology (Right).
4.5. Definition of the numerical calculation model
While the geometric aspect has been completely improved by the methodology presented, the
material characterizatión(homogeneus and isotropic), since at the time no experimental test were
carried out, the input loads and boundary conditions remain the same than the initial one. For
the loads have been consider: (i) Gravitational loads; (ii) Truss self-weight and (iii) Roof self-
weight.
Complementary to this loads conditions, it is necessary to correctly simulate the elastic
behaviour (Winkler model) of the ground in which the structure stands and also a proper
simulation of the behaviour of the tanscept. Such behaviour has been emulated through
CONTACT173/TARGET170 elements [40].
The discretization of the model has been carried out considering a 4-node isoparametric
tetrahedral element (SOLID65) with a maximum size of 0.60 m. In order to increase the
robustness of the tetrahedral mesh, element softening has been carried out using Laplace
algorithm [40].
All this results in a total of 218244 discrete elements into the numerical model (212537
SOLID65, 5707 CONTACT173 /TARGET170) (Fig.10).
Fig. 10. Isometric view of the mesh model (Left). Mesh detail of the balcony and chapels
(Right).
4.6. Modal analysis, calibration of the numerical model
The next step required to obtain an accurate finite element model is to calibrate their elastic
parameters and thus adapt the dynamic behaviour of the numerical model to the real one.In
order to accurately calibrate the numerical model it is necessary to follow a three-step
procedure: (i) Initial hypothesis, (ii) Manual calibration and (iii) Robust calibration. The initial
hypothesis considered were: the elastic propierties of masonry, the major cracks on the main
façade and the main nave, the elastic behaviour of the soil, and the simulation of the connection
between the nave and the transcept.
Within the manual calibration stage, numerous tests have been required, evaluating separately
each of the considered elastic variables and rejecting those that did not bring improvements to
the numerical model. Finally, have been chosen a total of eight parameters, namely: Young
modulu´s of the masonry (EMASONRY) and its density (δMASONRY),the normal (KNFAÇADE) and shear
(KTFAÇADE) stiffness of the major cracks on the main façade,the normal (KNFIRSTCAP) stiffness of
the major cracks on the main nave, the soil´s normal stiffness (KNSOIL), as well as the simulation
of the connection between the main nave and the transcept through a normal stiffness
(KNTRANSCEPT) and a shear one (KTTRANSCEPT). During the last step, the accurate calibration of the
previously discretized parameters through DR methodology (described above) was required,
providing the results presented below (Table 3).
Table 3: Summary of the adopted values for the calibration of the numerical model.
The now calibrated elastic properties of the masonry show a globally damaged material. The
high elasticity obtained in the simulation of the transept confirmed that the building, as it
progressed, suffered settlements and cracking. However, the rise in rigidity of the elastic
properties in comparison to the initially calculated in [6], denote soil compaction.As shown in
(Fig. 10) (Fig. 11), the relative ratios between the experimental and numerical frequencies, with
Initial values Lower bound Upper bound Updated value
EMASONRY (GPa) 10 5 15 9,19
δMASONRY (Kg/m3) 2500 2400 2700 2600
KNFAÇADE (GPa/m) 0.0001 0.00005 0.01 0.0004
KTFAÇADE (GPa/m) 0.1 0.05 1 0.53
KNFIRSTCAP (GPa/m) 1 0.05 5 0.40
KNSOIL (GPa/m) 0.585 0.0585 5.85 0.627
KNTRANSCEPT (GPa/m) 0.1 0.005 1 0.29
KTRANSCEPT (GPa/m) 0.0001 0.00001 0.01 0.00002
a value close to one for the standard MAC indexes denote a high correlation between the
experimental dynamic behaviour of the building and the numerical one. Through the simulation
of the cracks, it was possible to correctly simulate the dynamic behaviour and the high existing
elasticity in the central area of the building.
Fig. 10. Comparison between representative values of dynamic response (frequencies and
MAC): Value ratio of the initial model proposed by Lourenço (Left). Ratios obtained with the
hereby proposed methodology (Right).
Fig. 11. Comparison between the vibrational model from the OMA and those from the hereby
presented finite element numerical model.
5. Conclusions
The methodology presented aims to compile the information from different sensors in order to
establish a complete geometric characterization of historic buildings. The combined method for
data acquisition solves most common problems encountered today like the preparation of
accurate CAD models and the analysis of structure characteristics (displacements and cracking).
Using the laser scanner alone would not solve some of the problems that arise today, i.e. the
lack of data in areas that are not accessible to the system or the difficulty of cracking
identification. That is why it is essential to incorporate a complementary data capture system
able to meet solving problems. The perfect complement for an accurate, quick and complete
data capture is the Image Structure from Motion System on UAV platform. Other advantages
provided by this second capture system, lies in the potential that digital image analysis gives to
the graphical product. This digital image analysis makes possible to completely characterize
cracking (length and opening).
The binomial SfM-laser scanner is a reliable foundation from which to analyze appropriate
restoration actions, following the procedure: (i) analysis of the causes through the numerical
model (ii) displacement and stress state along the structure (iii) analysis of the effectiveness of
the system (analysis of the numerical model including restoration activities, analysis of cracks
and collapses systems).
The methodology presented can be applied to other infrastructures, such as tunnels or bridges,
given the high versatility of the sensors described and the wide range of possibilities that they
offer. All this is complemented by a global dynamic analysis of the structure that allows a
reliable calibration of the numerical model through the elastic system variables.
The Saint Torcato Church (Guimarães, Portugal) represents an ideal case study for evaluating
the potential of the method developed. Based on the geometric characterization performed
through the presented methodology, several research works are taking place, in order to improve
the characterization of the different materials by Ground Penetration Radar and Boroscopic
Camera, for further FEM analysis.
References
[1] Ramos LF, Aguilar R, Lourenço PB, Moreira S. Dynamic structural health monitoring of Saint
Torcato church. Mechanical Systems and Signal Processing. 2013;35(1–2):1-15.
[2] Grimm CT. Masonry Cracks: A Review of the literature. 1988:257-80.
[3] ICOMOS. Recommendations for the analysis, conservation and structural restoration of
architectural heritage. International Scientific Committee for Analysis and Restoration of
Structures of Architecural Heritage. 2003.
[4] Vercher J, Gil E, Mas A, Lerma C. Diagnosis/Intervetion creteria in damaged slabs by severe
corrosion of prestressed joists. Joirnal of Performance of Constructed Facilites. 2013.
[5] Betti M, Vignoli A. Assessment of seismic resistance of a basilica-type church under
earthquake loading: Modelling and analysis. Advances in Engineering Software.
2008;39(4):258-83.
[6] Lourenço PB, Ramos LF. Investiçasão sobre as Patologias do Santuário de São Torcato. Final
Report. University of Minho. 1999;99-DEC/E-5.
[7] Taliercio A, Binda L. The Basilica of San Vitale in Ravenna: Investigation on the current
structural faults and their mid-term evolution. Journal of Cultural Heritage. 2007;8(2):99-118.
[8] Armesto J, Roca-Pardiñas J, Lorenzo H, Arias P. Modelling masonry arches shape using
terrestrial laser scanning data and nonparametric methods. Engineering Structures.
2010;32(2):607-15.
[9] Pieraccini M, Dei D, Betti M, Bartoli G, Tucci G, Guardini N. Dynamic identification of
historic masonry towers through an expeditious and no-contact approach: Application to the
“Torre del Mangia” in Siena (Italy). Journal of Cultural Heritage. 2014;15(3):275-82.
[10] Riveiro B, Solla M, de Arteaga I, Arias P, Morer P. A novel approach to evaluate masonry
arch stability on the basis of limit analysis theory and non-destructive geometric
characterization. Automation in Construction. 2013;31(0):140-8.
[11] Lubowiecka I, Armesto J, Arias P, Lorenzo H. Historic bridge modelling using laser
scanning, ground penetrating radar and finite element methods in the context of structural
dynamics. Engineering Structures. 2009;31(11):2667-76.
[12] Park HS, Lee HM, Adeli H, Lee I. A New Approach for Health Monitoring of Structures:
Terrestrial Laser Scanning. Comp-Aided Civil and Infrastruct Engineering. 2007;22(1):19-30.
[13] Guidi G, Frischer B, Russo M, Spinetti A, Carosso L, Micoli LL. Three-dimensional
acquisition of large and detailed cultural heritage objects. Machine Vision and Applications.
2006;17(6):349-60.
[14] Yastikli N. Documentation of cultural heritage using digital photogrammetry and laser
scanning. Journal of Cultural Heritage. 2007;8(4):423-7.
[15] Pfeifer N, Briese C. Laser scanning: principles and applications summary. Third
international exhibition and scientific congress on geodesy, mapping, geology, geophysics,
Novosibrisk,Russia. 2007.
[16] Roca D, Lagüela S, Díaz-Vilariño L, Armesto J, Arias P. Low-cost aerial unit for outdoor
inspection of building façades. Automation in Construction. 2013;36(0):128-35.
[17] Barazzetti L, Binda L, Scaioni M, Taranto P. Photogrammetric survey of complex
geometries with low-cost software: Application to the ‘G1′ temple in Myson, Vietnam. Journal
of Cultural Heritage. 2011;12(3):253-62.
[18] TAPEnADe. Case Study:Maya bass-relief,Copan(Honduras). 2012.
[19] Westoby MJ, Brasington J, Glasser NF, Hambrey MJ, Reynolds JM. ‘Structure-from-Motion’
photogrammetry: A low-cost, effective tool for geoscience applications. Geomorphology.
2012;179(0):300-14.
[20] Besl PJ, McKay ND. A Method for Registration of 3-D Shapes. IEEE Trans Pattern Anal Mach
Intell. 1992;14(2):239-56.
[21] Toldo R, Beinat A, Crosilla F. Global registration of multiple point clouds embedding the
Generalized Procrustes Analysis into an ICP framework. 3DPVT 2010 Conference2010.
[22] Tognaccini R. La Chiesa di Santa Maria del Mar a Barcelona-Dal Rilievo Tridimensionale
all'Analisi Strutturale. Pisa,Italy: University of Pisa; 2009.
[23] Nguyen V-S, Bac A, Daniel M. Simplification of 3D Point Clouds sampled from Elevation
Surfaces. 2013.
[24] Varady T. Automatic procedures to create CAD models from measured data. Computer-
Aided Design and Applications. 2008;5(5):577-88.
[25] González-Jorge H, Gonzalez-Aguilera D, Rodriguez-Gonzalvez P, Arias P. Monitoring
biological crusts in civil engineering structures using intensity data from terrestrial laser
scanners. Construction and Building Materials. 2012;31(0):119-28.
[26] Armesto-González J, Riveiro-Rodríguez B, González-Aguilera D, Rivas-Brea MT. Terrestrial
laser scanning intensity data applied to damage detection for historical buildings. Journal of
Archaeological Science. 2010;37(12):3037-47.
[27] Barazzetti L, Scaioni M. Crack measurement: Development, testing and applications of an
automatic image-based algorithm. ISPRS Journal of Photogrammetry and Remote Sensing.
2009;64(3):285-96.
[28] Lubowiecka I, Arias P, Riveiro B, Solla M. Multidisciplinary approach to the assessment of
historic structures based on the case of a masonry bridge in Galicia (Spain). Computers &
Structures. 2011;89(17–18):1615-27.
[29] Deseilligny MP, DeLuca L, Remondino F. Automated Image-Based procedures for accurate
artifacts 3D modeling and orthoimage generation. Icomos. 2012;113.
[30] Friswell MI, Mottershead JE, Ahmadian H. Finite–element model updating using
experimental test data: parametrization and regularization. Philosophical Transactions of the
Royal Society of London Series A: Mathematical, Physical and
Engineering Sciences. 2009;359(1778):169-86.
[31] Ramos LF, De Roeck G, Lourenço PB, Campos-Costa A. Damage identification on arched
masonry structures using ambient and random impact vibrations. Engineering Structures.
2010;32(1):146-62.
[32] Mottershead JE, Friswell MI. Model Updating In Structural Dynamics: A Survey. Journal of
Sound and Vibration. 1993;167(2):347-75.
[33] Allemang RJ, Brown DL. A Correlation Coefficient for Modal Vector Analysis1982.
[34] Gentile C, Saisi A. Ambient vibration testing of historic masonry towers for structural
identification and damage assessment. Construction and Building Materials. 2007;21(6):1311-
21.
[35] Douglas BM, Reid WH. Dynamic Test and System Identification of Bridges. Journal of the
Structural Division. 1982;108(10):2295-312.
[36] Scaioni M, Barrazzetti L, Brumana R, Cuca B, Fassi F, Prandi F. Rc-Heli and Structure from
Motion techniques for the 3-D reconstuction of a Milan dome spire. ISPRS. 2009;XXXVIII-5/W1.
[37] Gordon S, Lichti D, Stewart M, Franke J. Structural Deformation Measuremente Using
Terrestrial Laser Scanners. 11th FIG Symposiom on Deformation Measurements, Santorini,
Greece. 2003.
[38] Lichti D, Gordon S. Error Propagation in Directly Georeferenced Terrestrial Laser Scanner
Points Clouds for Cultural Heritage Recording. FIG Working Week,Athens,Greece. 2004.
[39] Schaer P, Skaloud J, Landtwing S, Legat K. Accuracy Estimation for Laser Point Cloud
Including Scanning Geometry. 5th International Symposium on Mobile Mapping Technology.
Padova (Italy)2007.
[40] ANSYS. ANSYS Manual Sets. USA2012.